OBJECTIVES:To evaluate the role of partial recording protocols (PRPs) in reporting prevalence and severity of dental fluorosis and assess whether prevalence/severity estimates derived from PRPs differ by race/ethnicity. METHODS:Data from the National Health and Nutrition Examination Survey (NHANES) for the years 1999-2004 were analyzed with Stata(®) v.11. Prevalence of dental fluorosis obtained from a full-mouth examination (28 teeth gold standard) was compared with estimates derived from four subsets of teeth (maxillary canine-to-canine; maxillary first-premolar-to-first-premolar; all-premolars; all-molars). Sensitivity, negative predictive value (NPV), absolute bias, and correction factors were calculated against gold standard estimate. Analysis was stratified according to race/ethnicity to assess differences in estimates derived from PRPs. RESULTS:All subsets underestimated prevalence albeit to varying degrees. Two subsets (all-premolars and all-molars) had prevalence and severity estimates closest to gold standard estimates. The all-molars subset (eight teeth) recorded the highest sensitivity (84.5%) and the lowest absolute bias (3.5%) of all subsets relative to gold standard. Subsets derived from esthetically relevant teeth produced the lowest fluorosis prevalence. For instance, the maxillary canine-to-canine subset underestimated prevalence by 9.5%; incorporating the maxillary first premolars in the span improved prevalence estimate by 31%. Among non-Hispanic Whites, the all-premolars subset produced estimates closest to gold standard while the all-molars subset produced estimates closest to the gold standard among non-Hispanic Blacks and Hispanics. CONCLUSION:While the majority of dental fluorosis in the United States is very mild, concerns regarding its growing prevalence underscore the need for careful monitoring. The use of PRPs offers an alternative method of assessment, with validity of reported prevalence and severity dependent on choice of subset.